Andrew Nowicki <andrew@nospam.com> writes:

There is yet another option: a phased array
transmitter which makes only one narrow
microwave beam and changes its direction
every millisecond or so.

 ^^^^^^^^^^^^^^^^^
This would result in a signal bandwidth which was about 4 orders of 
magnitude sub-optimal for a SETI beacon.

In addition, signals are easier to detect if they can be integrated over
a reasonable observation period.  You can do this with pulsed signals - this
is what the pulse mode in the S@H client does - but it means that you have
to search over pulse repetition rates, as well as over frequencies.

If, as might be a reasonable assumption, pulses are phase coherent,
you can improve the sensitivity if you search over pulse lengths as well
as pulse durations and phase (so that you only look at the noise during
the pulses).  I believe S@H searches after coverting to scalar powers,
so doesn't do this.  However, the amount of time over which you can
do coherent processing of insterstellar signals is of the order of 20
seconds, and a fast pulsed signal will only put out a fraction of the
possible power in that period compared with a continuous signal or one
with pulses of much longer than 20s duration.